| Leaf area index, an important characteristic parameter of vegetation, is a reflection of the growth of the complex forest and the environment. It not noly can reflect the growth of vegetation, but also reflect the canopy structure of the vegetation by quantifying the indicators. Southwest evergreen broadleaf forest locates in the famous heavy rainfall areas of Sichuan, it has abundant rainfall, and is significant to the soil and water conservation, in addition to, it is an important part of the area in the upper reaches of the Yangtze River ecological barrier. Compared to the Age Plantations, the stand structure and composition of the natural evergreen broadleaf forest is much more complex, the forest ecosystem stable, the self-regulation and anti-interference ability strong, a larger effect on regional ecological benefits. With the rapid development of remote sensing technology, the use of satellite remote sensing technology to large regional studies LAI provides a good way. But over the years, scholars have studied less on Southwest evergreen broadleaf forest, and more focused on the study of a single scale surface parameters related to vegetation, is still relatively small in the multi-scale research.In this study, the research object is the Yucheng District Shangli town evergreen broadleaf forest, Pleiades-1, SPOT-5, Landsat-8 images based on ground truth combined with leaf area index (LAI) hemisphere photographic method obtained by partial least multiplicative regression, evergreen broadleaf forest at 2m,10m,30m scale LAI quantitative estimation model, Fitting LAI maps of the study area on three scales, and estimates for different models obtained were analyzed.(1) Combined with supervised classification and unsupervised classification of Pleiades-1, SPOT-5, Landsat-8 image, classification accuracy were 86%,82%,77.2%.(2) Based on multi-spectral image, using partial least squares regression establish LAI estimation model.a. Estimated LAI based Pleiades-1 images, by VIP importance of analyzing, selecting the six influence factors involved in modeling. Model: Y=0.51087* NDVI+0.34059* SAVI+0.06812*RVI+0.00012* DVI+0.00017* PVI+ 0.00006* NIR+1.84675, Overall accuracy was 86.5%.b. Estimated LAI based SPOT-5 images, by VIP importance of analyzing, selecting the six influence factors involved in modeling. Model:Y= 1.40863* NDVI+0.93965* SAVI+0.10733* RVI+0.0001* SWIR-0.00087* Red-0.00048* MR+1.0881, Overall accuracy was 81.4%.c. Estimated LAI based Landsat-8 images, by VIP importance of analyzing, selecting the night influence factors involved in modeling. Model: Y=-0.0001* NIR-0.00002* SWIR1-0.00007* SWIR2-0.00001* DVI+2.5539* NDVI-0.00002* PVI-0.1141* RVI+1.7022* SAVI-1.4696* SF+1.77242, Overall accuracy was 79.8%.(3) The writer compares the pros and cons of three kinds of images in terms of LAI inversion based on a more detailed analysis in terms of sensors, spectral resolution, band information, vegetation index in this paper. It has a good advantage to studing in a wide range of vegetation parameters LAI etc. based on Landsat-8 imaging; while the higher spatial resolution, the higher LAI retrieval accuracy, that is, in terms of small-scale accuracy LAI inversion, Pleiades-1 image has a great advantage; the higher spectral resolution, single-band surface information reflects the more accurate image information is also reflected in the more abundant.(4) By comparing this three distribution of leaf area index, has been found LAI distribution was the more broken which is based on Pleiades-1 image, but LAI maps which is based on Landsat-8 images obtained relatively continuous, LAI distribution which is based on SPOT-5 images between the two, which has a direct relationship with the retrieval of LAI scale. With the rising scale, the increase of the proportion of mixed pixels in the image lead to spectral information reflecting the feature type is large errors, and impact the inversion of vegetation parameters. |
PDF Full Text Request